Model averaging to estimate rebuilding targets for overfished stocks

Reducing overfishing and recovering overfished fish stocks is a challenging and important global problem. Rebuilding targets are essential ingredients for guiding overfished stocks to recovery. Having robust estimates of rebuilding targets is likely a necessary condition for effective long-term management of fishery resources. In this paper, we show how Bayesian model averaging can be applied to estimate rebuilding targets under alternative hypotheses about stock-recruitment dynamics. Using alternative hypotheses about stock-recruitment dynamics and environmental variation is important because using only a single hypothesis can lead to substantially different reference points and policy implications. The alternative hypotheses are augmented with prior information collected from meta-analyses of stock-recruitment data to construct a set of age-structured production models. We illustrate our approach using three overfished New England groundfish stocks. We fit alternative model likelihoods to observed data using Bayesian inference techniques. The Schwarz goodness-of-fit criterion was used to calculate model probabilities. Bayesian model averaging was used to estimate rebuilding targets that were robust to model selection uncertainty. Model-averaged estimates suggested that rebuilding targets for overfished stocks can be reasonably well determined if adequate prior information on stock productivity is available. Nonetheless, results had wide confidence intervals that reflected the underlying uncertainty in rebuilding targets.